John M Neff1, Holly Clifton2, Jean Popalisky3, Chuan Zhou4. 1. Center for Child Health, Behavior, and Development, Seattle Children's Hospital, Seattle, Wash. Electronic address: john.neff@seattlechildrens.org. 2. Department of Knowledge Management, Seattle Children's Hospital, Seattle, Wash. 3. Department of Clinical Effectiveness, Seattle Children's Hospital, Seattle, Wash. 4. Center for Child Health, Behavior and Development, Seattle Children's Research Institute, Seattle Children's Hospital, Seattle, Wash.
Abstract
OBJECTIVE: To stratify children using available software, Clinical Risk Groups (CRGs), in a tertiary children's hospital, Seattle Children's Hospital (SCH), and a state's Medicaid claims data, Washington State (WSM), into 3 condition groups: complex chronic disease (C-CD); noncomplex chronic disease (NC-CD), and nonchronic disease (NC). METHODS: A panel of pediatricians developed consensus definitions for children with C-CD, NC-CD, and NC. Using electronic medical record review and expert consensus, a gold standard population of 700 children was identified and placed into 1 the 3 groups: 350 C-CD, 100 NC-CD, and 250 NC. CRGs v1.9 stratified the 700 children into the condition groups using 3 years of WSM and SCH encounter data (2008-2010). WSM data included encounters/claims for all sites of care. SCH data included only inpatient, emergency department, and day surgery claims. RESULTS: A total of 678 of 700 children identified in SCH data were matched in WSM data. CRGs demonstrated good to excellent specificity in correctly classifying all 3 groups in SCH and WSM data; C-CD in SCH (94.3%) and in WSM (91.1%); NC-CD in SCH (88.2%) and in WSM (83.7%); and NC in SCH (84.9%) and in WSM (94.6%). There was good to excellent sensitivity for C-CD in SCH (75.4%) and in WSM (82.1%) and for NC in SCH (98.4%) and in WSM (81.1%). CRGs demonstrated poor sensitivity for NC-CD in SCH (31.0%) and WSM (58.0%). Reasons for poor sensitivity in NC-CD are explored. CONCLUSIONS: CRGs can be used to stratify children receiving care at a tertiary care hospital according to complexity in both hospital and Medicaid administrative data. This method will enhance reporting of health-related outcome data.
OBJECTIVE: To stratify children using available software, Clinical Risk Groups (CRGs), in a tertiary children's hospital, Seattle Children's Hospital (SCH), and a state's Medicaid claims data, Washington State (WSM), into 3 condition groups: complex chronic disease (C-CD); noncomplex chronic disease (NC-CD), and nonchronic disease (NC). METHODS: A panel of pediatricians developed consensus definitions for children with C-CD, NC-CD, and NC. Using electronic medical record review and expert consensus, a gold standard population of 700 children was identified and placed into 1 the 3 groups: 350 C-CD, 100 NC-CD, and 250 NC. CRGs v1.9 stratified the 700 children into the condition groups using 3 years of WSM and SCH encounter data (2008-2010). WSM data included encounters/claims for all sites of care. SCH data included only inpatient, emergency department, and day surgery claims. RESULTS: A total of 678 of 700 children identified in SCH data were matched in WSM data. CRGs demonstrated good to excellent specificity in correctly classifying all 3 groups in SCH and WSM data; C-CD in SCH (94.3%) and in WSM (91.1%); NC-CD in SCH (88.2%) and in WSM (83.7%); and NC in SCH (84.9%) and in WSM (94.6%). There was good to excellent sensitivity for C-CD in SCH (75.4%) and in WSM (82.1%) and for NC in SCH (98.4%) and in WSM (81.1%). CRGs demonstrated poor sensitivity for NC-CD in SCH (31.0%) and WSM (58.0%). Reasons for poor sensitivity in NC-CD are explored. CONCLUSIONS: CRGs can be used to stratify children receiving care at a tertiary care hospital according to complexity in both hospital and Medicaid administrative data. This method will enhance reporting of health-related outcome data.
Authors: Erik R Hoefgen; Annie L Andrews; Troy Richardson; Matthew Hall; John M Neff; Michelle L Macy; Jessica L Bettenhausen; Samir S Shah; Katherine A Auger Journal: Pediatrics Date: 2017-08-01 Impact factor: 7.124